About VGG16 pre-trained on ImageNet

[pengjw23]

we found that in the paper Chapter 4.2 : "ResNet101 [13] or VGG16 [36] pre-trained on ImageNet [7]". However, at adaptive_teacher/configs/faster_rcnn_VGG_cross_city.yaml, VGG16 did not used the pre-train ImageNet parameters like adaptive_teacher/configs/faster_rcnn_R101_cross_water.yaml.

We would like to know whether VGG16 are pretrained on ImageNet or not. thank you very much

[yujheli]

Good question. Our model using VGG16 is trained from scratch.

[pengjw23]

Thank you for your reply, and I want to know something about the dataset Foggy Cityscapes. I want to know which level weather you use when training, or all three types of weather are used for training. As a newcomer to this field, I am confused about this problem. I noticed that some papers only use foggy weather with level=0.02 for evaluation, for example, in Table 1 in PT (ICML-22, [1]).

[1]Chen M, Chen W, Yang S, et al. Learning Domain Adaptive Object Detection with Probabilistic Teacher[C]//International Conference on Machine Learning. PMLR, 2022: 3040-3055.

[Manjuphoenix]

@yujheli

Good question. Our model using VGG16 is trained from scratch.

So the parameter DATASETS.TRAIN inside Base-RCNN-C4.yaml file has no effect in the training process or does it use faster_RCNN model use the model pretrained on coco_2017_train dataset that is registered in the detectron?